Immigration, Latinos, and White Partisan Politics: The New Democratic Defection. Zoltan Hajnal, UCSD. Michael Rivera, UCSD.

Transcription

1 Immigration, Latinos, and White Partisan Politics: The New Democratic Defection Zoltan Hajnal, UCSD Michael Rivera, UCSD Abstract Immigration is profoundly changing the racial demographics of America. In this article, we seek to understand if and how immigration and increasing racial diversity are shaping the partisan politics of individual white Americans. We show that whites views on immigration and Latinos are strongly related to their core political identities and vote choices. Using a range of different surveys, we find that, all else equal, whites with more anti-immigrant views or more negative views of Latinos are less apt to identify as Democrats and less likely to favor Democratic candidates. This rightward shift harkens back to an earlier period of white defection from the Democratic Party and highlights the enduring but shifting impact of race on American politics. WORD COUNT: 9864

2 Immigration is transforming the demographics of America. In the last half century, the U.S. has become more diverse, Latinos have surpassed African Americans as the largest minority, and the proportion of the country that is white has fallen from roughly 90 percent to 65 percent. The future is likely to bring even more change. The Census projects that by sometime in the middle of this century, the U.S. will no longer be a majority white nation. In this article, we investigate the extent of the influence of immigration and racial diversity on the core political identities and voting preferences of individual white Americans. When white Americans choose to align with one of the two major parties, when they decide which candidate to support in Presidential contests, and when they vote in a range of other elections, do attitudes about immigration and Latinos help shape the outcome? Although widespread attention has been paid to the causes of our attitudes about Latinos and immigration more broadly, little research has focused on the consequences of immigrantrelated views (on causes see Schildkraut 2011, Hainmueller and Hiscox 2010, Kinder and Kam 2012, Brader et al 2008, Scheve and Slaughter 2001, Citrin et al 1997 but see Hopkins 2010). To date there is almost no direct evidence that the basic choices of individual white voters in American politics strongly reflect their views on immigration or the Latino population. Major recent studies of the presidential and Congressional vote tend to fall into one of two categories. Either they ignore immigration and race (McCarthy et al 2007 Miller and Shanks 1996, Alvarez and Nagler 1995, 1998), or if they focus on race, they limit that focus to the impact of America s old black-white divide (Lewis-Beck et al 2010, Valentino and Sears 2005, Abramowitz 1994, Carmines and Stimson 1989). No study that we know of has demonstrated a connection between 1

3 immigration and the white vote in national contests or revealed a link between immigration and white partisanship. 1 Nevertheless, we believe that immigration and the Latino population do impact whites electoral calculus and we offer an account of how large scale immigration can have real partisan consequences for the white population. First, immigration and the rapid growth of the Latino population have dramatically altered the racial group membership and imagery of the Democratic Party. Further, we believe that an oft repeated Latino (or immigrant) threat narrative has fueled individual white fears and insecurities about Latinos. Finally, when Republican and Democratic leaders take increasingly divergent stances on immigration, the two parties present individual white Americans with a compelling partisan logic. For those who are concerned about the Latino population, there is a powerful motivation to choose the Republican Party. Does Immigration Matter? There is incontrovertible evidence that race has mattered in American politics at different times in our history (Klinker and Smith 1999). And there is evidence that race still matters in American politics. Studies contend that whites policy preferences on welfare, education, crime, and a host of other cores issue are shaped by attitudes toward blacks (Gilens 2001, Hurwitz and Peffley 1997, Kinder and Sanders 1996 but see Sniderman and Carmines 1997). More critically, for our study, scholars have also linked partisan choices with racial attitudes. Several studies assert that whites defected from the Democratic Party in the 1960s in response to the Civil Rights Movement, the increased political participation of African Americans, and growing black support of the Democratic Party (Carmines and Stimson 1989, Huckfledt and Kohfeld 1989, Giles and Evans 1994). As blacks joined the Democratic Party in large numbers and as the 1 One important exception is work on California which alternately shows that Prop 187 led to growing white support of the Democratic Party (Bowler et al 2006) or that the episode had no impact on white partisanship (Dyck et al 2012). 2

4 Democratic and Republican Parties diverged on the main racial policy questions of the day, white identification with the Democratic Party especially in the South sharply declined. According to this view, whites sentiments about blacks helped Republicans dominate national elections (Valentino and Sears 2005, Edsall and Edsall 1991). And more recently racial views had, by many accounts, a substantial impact on Barack Obama s presidential bid (Lewis-Beck et al 2010, Bobo and Dawson 2009, Tesler and Sears 2010 but see Ansolabehere and Stewart 2009). There are, however, two concerns with this line of research. First, there are a number of authors who dispute just how much of this partisan shift was due to racial considerations (Abramowitz 1994, Lublin 2004). According to this view, other factors like social morality and more recently war, terrorism, and economic crisis have replaced race as the underlying basis for partisan choice (Adams 1997, Layman and Carmines 1997, Miller and Shanks 1996). If racial considerations do play an ongoing role in white partisan decision making, it is one that is questioned. Another concern with this research is that it focuses exclusively on the black-white divide while ignoring immigration and other racial dynamics. It is attitudes toward blacks and not views of Latinos or immigration that are purported to drive partisanship and the vote. However, given the dramatic growth in the Latino and immigrant populations, it is at least plausible that these groups have become more central in the political thinking of white America. Perhaps more significant for our purposes is the literature on minority context. Here an extensive set of studies has demonstrated the relevance of immigrant or Latino residential context for white Americans (Hopkins 2010, Ha and Oliver 2010, Hero and Preuhs 2007, Campbell et al 2006). But these findings are limited in one important way. Rather than look at the consequences of immigrant or Latino context for broad political outcomes like partisanship 3

5 and the vote, this literature tends to focus more narrowly on how immigrant or Latino context affects attitudes toward these minority groups (Ha and Oliver 2010,, Campbell et al 2006 ). 2 What research on the American case has not yet attempted to demonstrate is how immigrant context relates to the basic partisan choices of the white electorate. Comparative studies in Europe have identified clear links between the size of the national immigrant population and support for right-wing parties (Arzheimer 2009, Lubbers et al 2002). But the same has not been done in the United States. Ultimately, what is missing is compelling evidence that immigration is a core element of American politics. Why Immigration, Latinos, Party, and the Vote Are Linked But is there really reason to believe that immigration is driving major changes in white partisanship and the vote? From one, well-supported perspective, partisan identities are psychological attachments that are stable and generally impervious to change (Green et al 2002, Campbell et al 1960). Even Carmines and Stimson (1989) who so aptly demonstrated that an issue like race can dramatically alter the partisan landscape of the nation contend that most issues fail to generate significant partisan change. How can immigration overcome the inertia of partisan attachments? There are four different elements of immigration that we think make it a prime candidate to effect partisan evolution. First, immigration is salient a factor that is critical according to Carmines and Stimson (1989) and Carsey and Layman (2006). Americans tend to pay limited attention to the political sphere and their knowledge of the issues of the day is often quite limited, but immigration is not an ephemeral phenomenon that is easily overlooked (Delli 2 One important exception is an article by Hero and Preus (2007) that reveals a relationship between the size of the state immigrant population and welfare policy at the state level. 4

6 Carpini and Keeter 1996). Immigrants and their children now represent one in four Americans (U.S. Census 2005). The sheer size and racial diversity of the demographic change that has occurred and that continues to occur is impossible for white Americans to miss. Moreover, irrespective of the actual costs of immigration, there is an on-going and oft repeated threat narrative that links America s immigrant and Latino populations to a host of pernicious fiscal, social, and cultural consequences (Chavez 2008, Hopkins 2010, Santa Ana 2003). This narrative emphasizes cultural decline, immigrants use of welfare, health, and educational services, their propensity to turn to crime, and their tendency to displace native citizens from jobs (Huntington 2005, Borjas 2001, Gimpel and Skerry Each of these concerns has been spelled out repeatedly and in great detail in the media, in the political sphere, and in scholarly outlets (Valentino et al 2013, Chavez 2008, Perez N.D.). The result is that many Americans express real concerns about immigration. There is no doubt that views on immigration are diverse and that many Americans are either supportive or ambivalent about immigration. But it is also clear that many others are deeply concerned about immigration. Recent polls suggest that well over half of white Americans feel that immigrants are a burden on the nation, a slight majority think they add to the crime problem, about half believe they take jobs away from Americans, and perhaps most importantly an overwhelming majority of Americans view illegal as a serious problem (CNN 2010, USA Today 2010, CBS News 2010). Immigration is not generally viewed as the nation s most important problem but it is by almost all accounts a major phenomenon that produces real anxiety among large segments of the public. Immigration, like Carmines and Stimson s (1989) racial example and like other issues deemed to have caused shifts in partisanship, is also a relatively simple, symbolic issue. (Carmines and Stimson 1989, 1980, Layman and Carmines 1997, Adams 1997). There is considerable debate about exactly why Americans feel the way they do about immigration, but 5

7 studies suggest that attitudes toward immigration are linked to deep-seated, enduring predispositions like nativism, ethnocentrism and prejudice (Schildkraut 2011, Kinder and Kam 2012, Brader et al 2008, Burns and Gimpel 2000, Citrin et al 1990, 1997). How we think about Latinos, in particular, says a lot about our policy views on immigration (Valentino et al 2013, Brader et al 2008, Perez N.D.). As such, attitudes about immigration may be sufficiently deeply held and stable enough to sway partisan considerations. 3 Third and perhaps most importantly, immigration has done what few other issues can do. It has altered the group imagery associated with the two political parties. Even Green et al (2002) and Goren (2005) who write forcefully about the immovability and durability of party identification note that major shifts in partisanship have occurred over time as the social groups associated with each party have changed. With large scale immigration and growing Latino support of the Democratic Party, there seems little doubt that party images are changing. The vast majority of Latinos now favor Democratic candidates and the vast majority of Latino elected officials now represent the Democratic Party. 4 The growth of the immigrant population and the increasing attachment of Latinos and other immigrants to the Democratic Party means that a party that as late as 1980 was still 80 percent white is now more 40 percent non-white. 5 A party that was supported by lower class white interests increasingly became a party that was supported 3 At the same time, it is important to note that immigration diverges in important ways from the black-white divide that has often dominated American politics. Importantly, immigration tends to occur at higher rates in states where the black population is relatively small and thus may be threatening to new and different segment of the white population. Attitudes on immigration across the many surveys that we examine below are correlated with attitudes toward blacks but the two are far from synonymous. Correlations between the two typically range from insignificant to correlations of about.10 to.25. In short, not all Americans who oppose immigration have negative feelings toward African Americans. That means that immigration has the potential to be an important cross-cutting issue. Critically, until recent decades, Americans who expressed more anti-immigrant views were found equally in both parties. For example, in 1992 in the ANES Cumulative File, 41 percent of those white Americans who favored increasing the level of immigration identified the Republican Party, exactly the same percentage who identified with the Republican Party. 4 Latino Democratic identifiers now outnumber Latino Republican identifiers by a two-to-one margin (Hajnal and Lee 2011). Over two-thirds of Latino congress members and 84 percent of Latino state legislators are Democrats (NALEO 2013). 5 By contrast, roughly 90 percent of Republican identifiers are non-hispanic whites. Figures from ANES cumulative file and 2012 CCES. 6

8 by the black community and since the 1980s has increasingly become a party that is supported by Latinos and other immigrants. In other words, what it means to be a Democrat has changed. Finally, the two parties themselves have staked out increasingly divergent positions on the issue of immigration, another critical element in issue evolution (Carsey and Layman 2006, Carmines and Stimson 1989). Through the early 1990s, elites in the two parties were hardly distinct on immigration but that has slowly changed over time with partisan divisions first emerging at the state and local level and then expanding to Congress and finally to the Presidential level where in 2012 candidates from the two parties offered sharply contrasting positions on immigration for the first time. A range of empirical studies demonstrate this growing partisan divide in Congress (Wong 2013, Jeong et al 2011, Miller and Schofield 2008). There is also compelling evidence that Democratic and Republican leaders at the state and local level are increasingly divided on immigration (Ramakrishnan N.D). Partisan battle lines at the state level on immigration were most notably introduced in California when Governor Pete Wilson and the State Republican Party advocated for the Proposition 187, the so-called Save Our State initiative. The state level partisan divide has grown to the point that by 2010 no Democrat in the Arizona legislature supported the controversial immigrant enforcement bill, SB 1070, while all but one Republican voted for it (Archibold 2010). These divergent stances on immigration are borne out by interest group rates. Interest groups like Federation for American Immigration Reform (FAIR), the National Latino Congreso, and Numbers USA rate Democratic members of Congress as distinctly liberal on immigration and Republican members as strongly conservative. 6 When Republican leaders criticize immigrants, condemn their actions, and bemoan the costs to America, and when many 6 FAIR s ratings show little partisan divide on immigration as late as 1996 when Democrats averaged a score of 44 on immigration and Republicans received an average score of 52. But by the early 2000, FAIR s ratings by party sharply diverge. 7

9 Democratic leaders offer support for the plight of immigrants, they present citizens with a choice on an issue that many feel is threatening America. In short, many white Americans will see that America is changing, will believe that immigration is driving many of the negative changes they see, will know that one party is backed by immigrants and stands largely on the side of immigrants while the other party is opposed by almost all immigrants and stands largely in opposition to immigration. For many white Americans, this may be a powerful motivation to defect to the Republican Party. Research Design In order to assess the impact of immigration and immigration-related views on the politics of white America, we turn to a standard tool of American public opinion survey research the American National Election Study (ANES). We choose the ANES because it includes a long list of questions that get at each of the many different factors known to affect partisanship and the vote. This is critical, since we cannot know if immigration matters, unless we can control for all of the core aspects of American elections. We begin with an analysis of the 2008 ANES for two reasons. First, it contains questions on immigration a requirement that rules out most years of the ANES survey and many other surveys. Second, 2008 was ostensibly not about immigration. Barack Obama, the first African American nominee for President was on the ballot, McCain and Obama outlined similar plans on immigration, the nation was in the midst of two wars, and it faced an almost unprecedented fiscal crisis. Immigration was supposedly not a critical issue in the campaign. If anything, 2008 was going to be about whites acceptance of blacks and their concerns about the economy, war, and terrorism. As such, 2008 represents a relatively exacting test of our immigration hypothesis. We realize, however, that if we want to make a more general statement about American politics, we need to assess the influence views of Latinos and immigrants have across a wider 8

10 range of data sets, years, elections and contexts. To do this we repeat our analysis using the ANES cumulative file, the 2000 and 2004 National Annenberg Surveys (NAES), and the 2010 and 2012 Cooperative Congressional Election Study (CCES). This allows us to test the immigration hypothesis across different years (contests from ), different types of elections (President, House, Senate, Gubernatorial), different types of survey instruments (including a wide variety of questions that vary the wording of the key independent variable immigration related feelings and the key dependent variables partisanship and vote choice), and different survey methodologies and samples. Finally, since party choices may impact rather be impacted by immigrant related views, we undertake causality tests on the three ANES panel data sets that include questions on immigration. If all of these different data points lead to the same story, we can be reasonably confident of that story. Because our theory focuses on the reaction of white Americans to America s changing racial demographics, we include only those individuals who identify themselves as white and as non-hispanic. Defining and operationalizing our key independent variable views toward Latinos and immigration is not straightforward. The process is complex because we believe that white Americans tend to conflate several distinct categories of people. Although in theory categories like illegal immigrant, immigrant, and Latino are all distinct, in the practice and rhetoric of American politics these concepts often blur together (Perez N.D., Chavez 2008). 7 In light of these muddled categories, we will test a series of different measures of Latino and immigrant views to try to get a clearer sense of just who it is that white Americans are reacting to. 7 In reality, Latinos make up only a little over half of all foreign born Americans and undocumented immigrants represent only about 28 percent of all the foreign born (US Census Bureau 2010, Passel 2009). Nevertheless, the majority of Americans believe that most immigrants are illegal (Enos 2012) and studies show that the issue of immigration is strongly bound with one group Latinos. Brader et al (2008) and Valentino et al (2013) demonstrate that images of Latinos more than non-latinos elicits more opposition to immigration and Perez (N.D.) finds that implicit attitudes toward Latinos are highly predictive of opinions on immigration policy. 9

11 Since white Americans express the strongest reservations about illegal immigrants, we begin by focusing on a summary measure of views on illegal immigration. Specifically, we use the four questions in the 2008 ANES that explicitly address illegal immigration to create an Alpha factor score for each respondent. The scale comes directly from the factor analysis and ranges from -2.8 to 1.7 with higher values representing more positive views of immigration. The four questions are: 1) a standard feeling thermometer for illegal immigrants that ranges from 0 (meaning extremely negative feelings) to 100 (for extremely positive feelings), 2) Should controlling and reducing illegal immigration be a very important. not an important foreign policy goal?, 3) Do you favor/oppose the U.S. government making it possible for illegal immigrants to become U.S. citizens?, and 4) Do you favor, oppose, or neither favor nor oppose allowing illegal immigrants to work in the United States for up to three years after which they would have to go back to their home country? The four items cohere well with a scale reliability of.65 and an average inter-item correlation of.32. In practice, it matters little how we combine these questions or whether we focus on a subset of these questions or on just one of these questions. A simple additive scale performs similarly in the regressions that follow. Also, in alternate tests when we substitute each single question or combinations of two or three of these questions into the regressions, the pattern of results is similar. Since we think concerns about a range of different groups (immigrants, illegal immigrants, Latinos) are clustered together in the minds of many white Americans, we incorporate a range of different measures of feelings toward these groups into our tests. Specifically, in alternate tests of the 2008 ANES data we examine attitudes toward immigration in general ( Should immigration levels be increased a lot decreased a lot? ), and attitudes toward Latinos (a standard feeling thermometer for Hispanics ). The results of these alternate tests are described below. 10

12 Across the other public opinion surveys that we examine, questions on immigration vary substantially. Earlier and later versions of the ANES address whether immigration is a burden and include a standard feeling thermometer toward Hispanics. The NAES focuses on whether the federal government should do more to restrict immigration. And the CCES asks about spending on the border patrol. Despite the substantial variation in the content of these questions, there is a consistency of findings. In each case, Latino or immigrant-related views are significantly and substantively tied to partisan choices. The main focus of this study is on partisanship and the vote. Our main measure of partisanship is the standard 7 point party identification scale. Respondents place themselves on a scale that ranges from strongly Democratic (1) to strongly Republican (7). To assess the robustness of our results, in alternate tests, we also direct our attention to party feeling thermometers, dummy variables isolating Democratic identifiers and Republican identifiers, and unordered party identification models (utilizing multinomial logistic regressions). We assess the vote in as many ways as possible. The ANES has the vote for Presidential, Congressional, Senatorial, and Gubernatorial contests. In the 2008 Presidential contest we focus primarily on a simple dummy variable which indicates either a vote for the Democratic candidate (0) or a vote for the Republican candidate (1). In other contests with significant third party candidates, we utilize an unordered 3 point scale (Democrat, Republican, Third Party). With the ANES we can also assess the effects of Latino and immigrant related attitudes on intended vote choice and candidate feeling thermometers. One of the most difficult aspects of this empirical endeavor is ensuring that we include controls for all of the different factors that could drive white s electoral decisions and be correlated with white views on immigration (see Miller and Shanks 1996 for an overview of the partisan choice literature). In short, our empirical models have to incorporate key elements of 11

13 American politics. With that in mind, we include measures for: 1) basic ideology the standard seven point liberal-conservative self-placement scale; 2) military action support for wars in Iraq and Afghanistan and views on expanding the war on terrorism; 3) retrospective evaluations presidential approval and assessments of the economy; 4) redistribution higher taxes for the rich and welfare spending; 5) morality and religion - views on homosexuality and the importance of religion ; 6) views of blacks Kinder and Sanders (1996) four standard racial resentment questions; 7) other racial attitudes/ethnocentrism standard feeling thermometers for blacks, Asian Americans, and whites ; and 8) in alternate tests other issues like universal health care, women s rights, the environment, abortion, crime, schools, science and technology (see online appendix for question wording). 8 Also, since partisan choices have been linked to class, religion, and other individual demographic characteristics we control for education (number of years of school completed), household income (divided into 25 categories), gender, age in years, whether the respondent is unemployed or not, whether anyone in the household is a union member or not, marital status (married or not), and religious denomination (Jewish, Catholic, Protestant, or Other). In alternate tests, we also account for self-identified class status, church attendance, whether the respondent is born-again or not, and years living in the community. All told, we have controls for basic ideology, retrospective evaluations, a range of core issues, racial attitudes, and individual social characteristics many if not all of the factors that are presumed to dominate the vote. Views on Immigrants and Latinos and Partisanship 8 These other issues are not included in the main model because they are only asked of half of the respondents. 12

14 In Table One, we begin to assess the connection between immigrant or Latino related views and partisanship. 9 The table displays a series of regressions that control for an increasing number of factors from socio-demographics characteristics to issue positions, ideological views, and racial attitudes all purportedly central to partisan choice in America. Each model is an OLS regression with the standard seven point party identification scale as the dependent variable. Table One. Views on Immigration and White Partisanship 2008 ANES Party Identification (High=More Republican) Model 1 Model 2 Model 3 Model 4 IMMIGRATION Positive Views Toward Illegal Immigrants -.61 (.09)** -.24 (.09)** -.22 (.09)* -.19 (.09)* DEMOGRAPHICS Education -.01 (.03).02 (.03).01 (.03).01 (.03) Income.04 (.01)**.03 (.01)**.03 (.01)**.03 (.01)** Unemployed.16 (.33).22 (.29).25 (.29).29 (.29) Age -.08 (.04)* -.05 (.03) -.03 (.03) -.07 (.04) Female -.20 (.14).06 (.12).04 (.12).03 (.12) Married.57 (.14)*.14 (.12).14 (.11).16 (.12) Union Member -.53 (.21)* -.51 (.16)** -.52 (.17)** -.54 (.17)** Jewish (.52)* -.23 (.42) -.14 (.42) -.31 (.42) Catholic.16 (.18) -.09 (.16) -.05 (.16) -.11 (.16) Protestant.80 (.16).17 (.14).21 (.14).11 (.14) IDEOLOGY Conservative.62 (.05)**.60 (.05)**.61 (.05)** ISSUE POSITIONS War and Terrorism Expand War on Terror Support War in Iraq Support War in Afghanistan Economy/Retrospective Economy Improving Approve President Redistribution Favor Higher Taxes on Rich Increase Welfare Spending Morality/Religion Favor Gay Rights Religion Important Racial Resentment to Blacks.02 (.03) -.02 (.04) -.08 (.04)* -.10 (.09) -.43 (.05) -.14 (.06)* -.01 (.03) -.05 (.04).05 (.03).02 (.03) -.02 (.04) -.08 (.04)* -.09 (.09) -.42 (.05) -.12 (.06)* -.02 (.03) -.06 (.04).06 (.03).02 (.03) -.02 (.04) -.08 (.04)* -.10 (.09) -.43 (.05) -.15 (.09)* -.01 (.03) -.05 (.04).06 (.03) 9 Simpler, bi-variate tests show that views on immigrants are strongly and significantly correlated with not only partisanship (r=.22 p<.001) but also the Presidential vote (r=.33 p<.001), the Congressional vote (r=.30 p<.001) and the Senatorial vote (r=.30 p<.001). 13

15 Blacks Deserve Less Blacks Get Special Favors Little Discrimination Blacks Should Try Harder Other Racial Considerations Warmth Toward Blacks Warmth Toward Asians.17 (.07)*.12 (.06)* -.02 (.06).03 (.06) -.81 (.52) 1.18 (.52)*.59 (.43) Warmth Toward Whites Constant 4.39 (.50)** 3.36 (.69)** 2.46 (.74)** 3.36 (.69)** N Adj R Squared **P<.01 *P<.05 We start with a basic model that includes a traditional set of socioeconomic characteristics and other demographic variables. Given claims about class and religious based support for each party, we include basic markers of class (education, income, employment status, union membership), and a series of dummy variables measuring religious affiliation (McCarthy et al 2007, Adams 1997, Layman and Carmines 1997). Model 1 suggests that many of these measures are important for partisanship but more importantly, it shows that net basic demographic controls, attitudes toward illegal immigration are closely linked to partisan attachments. All else equal, non-hispanic whites who have more negative feelings toward illegal immigrants are predicted to be just over one point more Republican on the seven point party identification scale than are whites with less negative views. Given that a one point shift equals the difference between a strong Democrat and a weak Democrat, immigrant views could be greatly re-shaping American politics. 10 Political choices in America are obviously about much more than immigration or Latinos. There is little doubt that recent elections have focused significantly on America s ongoing economic recession, its wars in Iraq and Afghanistan, the enduring terrorist threat facing the nation, social morality issues like gay rights, and a core ideological dimension liberalism vs 10 For this and all other predicted probabilities reported in the paper, estimates were calculated using Clarify holding all other independent variables at their mean or modal value and varying the independent variable of interest plus or minus one standard deviation. 14

16 conservatism (Abramson et al 2007, Layman and Carmines 1997, Fiorina 1981, MacKuen et al 1989). In Model Two, we incorporate each of these different elements in our model of party identification. What we find confirms much of what we know about American politics. Most of these issues, the basic ideological orientation, and retrospective evaluations greatly influence which party individual Americans choose to support. What is striking, however, is that the inclusion of all of these different elements of American politics does not eliminate the impact of views on immigrants. Views of illegal immigrants still significantly shape white partisanship after controlling for a range of measures of issues, ideology, and retrospective evaluations. Moreover, alternate tests indicate that it does not matter which issues we include or how we measure issues, ideology, and retrospective evaluations. When policy questions on health care, crime, foreign aid, schools, women s rights, the environment, and science are added to the model, the impact of immigrant-related views on partisan attachments is largely unaffected. Further, immigrant-related views remain significant when we substitute in alternate measures of economic policy preferences or retrospective evaluations. No matter what one s views on the economy, the war, abortion, and other factors, views of illegal immigrants are strongly associated with being a Republican. Immigrants and Latinos or Blacks and Ethnocentrism? One element of American politics that we have largely ignored to this point is the blackwhite divide. When race has mattered in American national elections, the main issue has usually been the rights and interests of African Americans (Klinker and Smith 1999, Carmines and Stimson 1989, Key 1949). Especially in 2008, with Barack Obama, the nation s first black presidential nominee on the ballot, and evidence that racial resentment played a role in the white 15

17 vote, these kinds of racial attitudes need to be integrated into the analysis (Lewis-Beck et al 2010, Bobo and Dawson 2009, Tesler and Sears 2010 but see Ansolabehere and Stewart 2009). Thus, in Model Three, we add four different questions from the racial resentment scale developed by Kinder and Sanders (1996) and included in most bi-annual editions of the ANES. All four measures explicitly ask about attitudes toward African Americans and combined the four measures have been shown to play a critical role in white public opinion (Kinder and Sanders 1996). The results in Model 3 indicate that the black-white divide remains significant in white partisanship. Whites who are more racially resentful of blacks are predicted to be 1.1 points more Republican on the party identification scale than are whites who are less resentful of blacks. But the results also suggest that immigration represents a distinct dimension that helps to shape white partisan ties. Even after considering the effects of racial resentment toward blacks, those who have more negative views of illegal immigrants continue to be significantly more apt to identify as Republican. Attitudes on immigration are not merely proxies for racial attitudes. In the last model of Table 1, we further investigate the role of race and the possibility that immigrant-related views are a stand in for some deeper aspect of America s racial dynamics like racial prejudice or ethnocentrism (Kinder and Tam 2009, Brader et al 2008, Burns and Gimpel 2000). Specifically, we incorporate whites views of African Americans, their views of white Americans, and their views of Asian Americans. 11 In each case, we utilize a basic feeling thermometer toward each group. Despite the inclusion of feelings toward the three different racial groups in the model, we still find that immigrant-related views are important for white partisanship. Whites with the most negative views of illegal immigrants are predicted to be one 11 Since we believe that attitudes toward immigrants and Latinos are closely linked, we do not include a Latino feeling thermometer. 16

18 third of a point higher on the seven point party identification scale than are whites with the most positive views of illegal immigrants. The impact of immigration on American politics cannot be wholly reduced by incorporating traditional measures of stereotypes and ethnocentrism. To assess the robustness of these results, we repeated the tests in Table One with a range of different measures of immigrants and Latinos. Specifically, when we substituted in a measure of feelings toward Latinos (a Hispanic feeling thermometer), a measure of feelings toward legal immigration (should immigration levels be increased or decreased), and a simpler measure of feelings toward illegal immigrants (an illegal immigrant feeling thermometer), all were significant in the regression model. Regardless of how we measure attitudes toward immigrants or Latinos, these attitudes are closely connected to party identification. In alternate tests, we also assessed different party-based dependent variables. Specifically, we examined feeling thermometers toward each party, dummy variables for identity with each party, and an unordered three point party identification scale (multinomial logistic regression model). In each case, views toward illegal immigrants remained significant and the effects were generally substantial. For example, all else equal, those with more positive views of immigrants scored 6 points higher on the Democratic Party feeling thermometer. Regardless of how one measures partisanship, it appears to be closely linked to views on immigration. One concern with the analysis that we have presented is the possibility of reverse causation. It is possible that party identification may impact rather than be impacted by immigrant related views. Indeed, much of the literature in American politics suggests that party identification stands near the beginning of a funnel of causality that drives factors like issue positions (Campbell et al 1960 but see Dancey and Goren 2010). We do not dispute that party identification is the prime mover in American politics but we nevertheless maintain that deep seated attitudes on immigration and race can shift the partisan leaning of some members of the 17

19 population. We directly test that proposition in three ways. First, we preform causality tests on the ANES panel to see if pre-existing attitudes on immigration do, in fact, predict future changes in individual party. Second, later in the article we shift the focus to the vote rather than partisanship, where we find that immigrant views are correlated with vote choice after controlling for party identification. Third, and also later in the article, we look separately at Democrats, Independents, and Republicans to see if views on immigrants matter within each party. These later two tests allow us to see if views of Latinos or immigrants have a relationship with the vote that goes beyond partisanship. Finally, we look at aggregate data to see if lagged immigration views predict subsequent changes in white macropartisanship. For our first causality test we focus on the most recent ANES panel. The basic idea is to determine if past views on immigration predict current partisanship net the effects of past partisanship. In other words, do past views on immigration help predict future changes in party identification? As Table Two illustrates, there is a clear, temporal link between immigrant views and partisanship. Views on immigration (measured by a question about whether illegal immigrants should be given a chance to become citizens) in 2008 have a significant effect on party identification measured in 2009 after controlling for party identification measured in Indeed, even after controlling for past partisanship, a one standard deviation shift in views of illegal immigrants is tied to about a one quarter point shift on the 7 point party identification scale. Attitudes on immigration are not leading to a wholesale shift from strong Democrat to strong Republican over the course of a year but feelings about immigrants do appear to be leading to some very real changes in partisanship. 12 This is true whether we use the 2008 ANES panel or instead perform the test on the or ANES panels. Importantly, the 12 At the same time it is important to note that by the same test party identification does cause changes in immigrant related views. The relationship between party identification and immigrant-related views is reciprocal. 18

20 influence of immigrant-related views on partisanship persists even when we control for past partisanship and a range of other major issues typically linked to partisanship (see the online appendix). Table Two. Assessing Causality: Immigration s Temporal Impact on Party Identification in 2009 Party Identification (2008) High=Rep.89 (.01)** Views on Illegal Immigrants (2008) High=Pro Immig -.03 (.01)* Constant.21 (.06)** N 1171 F 2603 ** **P<.01 *P<.05 Immigrant Views and the Vote Are the right-ward shifts that we see on partisan attachments accompanied by a shift to the right in national electoral contests? To answer this question, we undertake logistic regressions that analyze the vote in the 2008 presidential election. The dependent variable in each case is a dummy variable indicating support for the Republican candidate - McCain (1) or support for the Democratic candidate - Obama (0). In one regression we focus on the reported vote of respondents queried after the election. In the second regression, we examine intended vote choice for those surveyed before election day. We include the same list of controls that we did earlier for party identification with one exception. Since we are particularly interested in determining if immigration has an effect on the vote that goes beyond party affiliation, we add the standard 7 point party identification scale to the list of controls. By including party identification in our vote models, we can conclude with some confidence that views on immigration have an independent effect that is not wholly driven by party identification. The regression results are displayed in the online appendix. Figure One illustrates the effect of immigration on the presidential vote net all of the controls. 19

21 FIGURE ONE HERE Figure One shows that how we think about immigrants is strongly related to the vote. As we saw before, whites with more negative attitudes toward illegal immigrants are significantly more likely to opt for Republican options. All else equal, more negative views of illegal immigrants are associated with a 23.7 percent increase in the probability of voting for John McCain, the Republican presidential candidate. The effect for intended vote choice is almost identical a 22.9 percent increase in the probability of voting for McCain. Impressively, in an election that occurred in the midst of one of the nation s sharpest recessions in history, that coincided with two wars in Afghanistan and Iraq, and that included the nation s first black presidential nominee, views on immigrants still mattered. 13 To further test the role of immigrant and Latino views on the vote, we assessed the impact of immigrant-related views within each party. By looking within party, we get another look at how attitudes toward immigration matter beyond partisanship. Among those who claim ties to the Democratic Party, views of illegal immigrants are significantly related to vote choice. The vast majority of Democrats vote for Obama but those who have more negative views of illegal immigrants are 6.5 percent less likely to vote for Obama than those with more positive views of illegal immigrants. This is a small sign that immigration is pushing whites Democrats away from their party. By contrast, the results indicate that views toward immigrants matter little for Republicans. This is, however, what we would expect to find if immigration is pushing whites in one direction toward the Republican Party. Also as one might expect, views toward immigrants and Latinos have the largest impact on non-partisans. White independents who hold 13 These results are robust to different ways of measuring the dependent variable. If we focus on feelings toward the Democratic and Republican candidates rather than on the vote itself, we once again find that more negative views of illegal immigrants are associated with stronger, more positive feelings for the Republican side and less positive views of the Democratic option. 20

22 more negative views of immigrants are 67.7 percent more likely to vote for McCain than white independents who hold more positive views of immigrants (analysis in the online appendix). Views of Latinos and Immigrants in Other Elections To make a general statement about the impact of immigration in American politics, we have to look more broadly at a number of different presidential elections as well as across a range of different types of electoral contests. This is exactly what we do in Table Three. Specifically, we turn to the ANES cumulative file to assess the impact of immigration views on Presidential, Congressional, Gubernatorial, and Senate contests. Since the ANES does not generally ask about views on illegal immigrants, we utilize a different measure for attitudes towards immigrants and Latinos. The key independent variable here is the standard feeling thermometer toward Hispanics. Also, since policy questions vary from ANES year to year, we include a modified set of policy control variables (see online appendix). Table Three. Views on Latinos and the Vote Alternate Measures ANES Cumulative File Presidential Vote Support for the Republican Candidate. (Multinomial Logit) Intended Pres Vote for Vote for Dem vs Ind Rep vs Ind Vote Congress Governor 21 Vote for Senate IMMIGRATION Views of Hispanics -.67 (.62) (.61)* (.45)* -.72 (.47) (1.35)**.23(.39) DEMOGRAPHICS Education.36 (.08)**.30 (.08)** -.07 (.07) -.07 (.07) -.03 (.15) -.01 (.06) Income.01 (.06).04 (.06).03 (.05).02 (.05) -.08 (.13).02 (.04) Unemployed -.06 (.37) -.44 (.40).12 (.34) -.45 (.38) -2.0 (1.4) -.38 (.34) Age.02 (.00)**.01 (.05) -.01 (.01) -.01 (.01) -.01 (.01) -.01 (.00)* Female.32 (.15)*.62 (.15)**.09 (.12).33 (.13)**.09 (.28).04 (.11) Married -.46 (.16)** -.11 (.16).39 (.13)**.30 (.13)*.32 (.30).16 (.11) Union Member.35 (.19) -.11 (.19) -.64 (.15)** -.56 (.15)** -.56 (.36) -.40 (.13)** Jewish.95 (.58).32 (.62) -.43 (.39) -.83 (.41)* -.51 (.90) -.90 (.37)** Catholic -.15 (.23).20 (.24).37 (.20).27 (.22).76 (.45).07 (.18) Protestant -.06 (.21).25 (.22).26 (.18).20 (.20) -.45 (.43) -.15(.11) IDEOLOGY/PARTY ID Conservative -.25 (.06)**.20 (.07)**.35 (.05)**.44 (.06)**.19 (.13).24 (.05)** Republican -.54 (.05)**.31 (.05)**.84 (.03)**.83 (.04)**.65 (.09)**.55 (.03)** ISSUE POSITIONS War and Terrorism More for Military.09 (.06).38 (.06)**.28 (.05)**.29 (.05)** -.01 (.11).14 (.04)**

23 Economy/Retrospective Economy Improving Approve President Role of Government More Govt Services -.45 (.08)**.13 (.08).25 (.06).00 (.05) -.48 (.08)** -.40 (.07)**.03 (.05) -.11 (.05)*.10 (.06) -.56 (.06)** -.22 (.05)** -.15 (.04)** -.03 (.06) -.51 (.06)** -.27 (.05)** -.13 (.04)**.22 (.15).24 (.15) -.01 (.11) -.17 (.09) -.01 (.05).07 (.05) -.11 (.04)** -.12 (.04)** Favor Guaranteed Job Social Issues Favor Women s Rights.00 (.06) -.15 (.05)** -.09 (.04)* -.15 (.04)** -.23 (.10)* -.14 (.04)** Racial Considerations Warmth toward Blacks Warmth toward Whites.37 (.63).92 (.52).45 (.62) 1.36 (.51)**.34 (.47).39 (.39).20 (.49).52 (.42) 2.79 (1.40)*.09 (.91) -.14 (.42) -.20 (.34) Constant -59 (24) -27 (24) 49 (18)** 36 (19) -4.1 (1.5)** N Adj R/ Pseudo Squared **P<.01 *P<.05 Our results suggest, once again, that how white Americans think about Latinos can be a central component of white Americans electoral calculations. Starting with the first two columns which display the results of a multinomial logistic regression with presidential vote choice Democrat, Independent, Republican as the dependent variable, we see that those who feel more warmly to Hispanics are significantly less apt to choose Republican candidates for president. The third column, which displays the results for intended presidential vote choice (with a Republican vote as the dummy dependent variable) reconfirms the results. Again, more positive views of Latinos are significantly tied to Republican vote choice net party identification and a range of other controls. Moreover, the magnitude of the relationship is substantial. A two standard deviation negative shift in view of Hispanics is associated with a 9.8 percent increase in the probability of Republican vote choice in the multinomial model. For intended vote choice, the comparable figure is a 10.9 percent increase in Republican voting. And for recalled vote from the last election (analysis not shown), the figure is an 8.9 percent gain in Republican vote 22

24 probability. Across a range of presidential elections no matter how we measure vote choice we see that attitudes toward Latinos are very much a part of vote choice. 14 Importantly, the relationship is not isolated to presidential vote choice. As the rest of the columns demonstrate, white views of Latinos are significantly linked to gubernatorial vote choice and almost significantly tied to the Congressional vote. Moreover, in gubernatorial contests, the magnitude of the relationship is large. All else equal, those who hold more negative views of Latinos are 35 percent more likely to favor Republican gubernatorial candidates. The one case where there is no apparent relationship is in Senatorial contests. 15 Robustness Checks To help ensure that the results to this point measure the underlying relationships between immigration related views and white partisan choices, we performed a serious of additional tests. First, we repeated as much of the analysis as possible with a number of different data sets. Using the 2000 and 2004 National Annenberg Election Surveys (NAES) we examined the link between immigrant related views (should the federal government do more to restrict immigration and is immigration a serious problem) and party identification in 2000 and 2004, vote choice in the 2000 and 2004 presidential elections, and intended vote choice in presidential (2000 and 2004), Senatorial (2000), and House elections (2000). In every case except for Senate elections, after controlling for a range of factors that were purported to drive electoral behavior in that year, 14 Interestingly, feelings toward Asian Americans do not have the same effect. Across the vote and partisanship tables, feelings toward Asian Americans more often than have no significant effects and in one case actually have positive effects on the Republican vote. Given that Asian Americans hold on average higher economic status than Latinos, given that stereotypes of Asian Americans are very different from stereotypes of Latinos (hardworking, intelligent, and foreign vs poor, violent, and less intelligent), given that Asian Americans represent a much smaller share of the population than Latinos, and finally given that Asian Americans have only very recently sided in large numbers with the Democratic Party, it is unlikely that the Asian American population will spark the same political reaction as the Latino population (Bobo et al 2000, Hajnal and Lee 2011). 15 We endeavored to see if there was any pattern over time or across space in the effects of immigration on partisanship and the vote. We could not find a clear and consistent pattern but speculate that the lack of clear results is due to data limitations (see online appendix). 23